Jennifer Marsico, Research Assistant, AEI-Brookings Election Reform Project
November 5, 2008

Passed in 2002, the Help America Vote Act (HAVA) sought to improve the administration of federal elections, and, in turn, to ensure that each person who casts a vote has that vote counted accurately. But in the state of New Jersey, though HAVA has fueled a push towards this goal, reality has fallen short of the ideal. HAVA's own requirements are partially to blame; as a December 2005 Congressional Research Service report (PDF) points out, the law's requirement that individuals with disabilities be able to cast an ballot independently has encouraged greater use of direct-recording electronic voting machines, or DREs. All of New Jersey's 21 counties now use DREs as their standard polling place equipment. The vast majority of which rely on full-face DREs, where the entire ballot appears on one screen, as opposed to scrolling or paging DREs, where the voters' choices are spread out over a sequence of screens. Among the counties utilizing full-face DREs, furthermore, most employ a single model: the Sequoia AVC Advantage. Although the use of DREs has allowed these counties to comply with HAVA, a range of past problems exhibited by DREs suggest that these machines should be looked at with a skeptical eye.

First and foremost, DRE failure is higher than that of other voting technologies-often, considerably higher. As Paul Herrnson and his co-authors note in their recently published book Voting Technology: The Not-So-Simple Act of Casting a Ballot, voting technology and ballot layouts can have an impact on the occurrence of "undervotes," or not voting for each item on the ballot. In another recent study, David C. Kimball of the University of Missouri-St. Louis and Martha Kropf of the University of North Carolina-Charlotte find that voting technologies can also have a significant impact on the residual vote rate (the percentage of ballots with invalid votes for a particular office). New Jersey's DREs have the benefit of making "overvotes" impossible, as they allow only one button to be pushed for each office or public question. Yet for full-face DREs, the residual vote rates due to undervoting can still be troublingly high. An October 2008 paper (PDF) argues that the Sequoia AVC Advantage does little to prevent high numbers of undervotes, as not only is the full-face ballot design "densely printed with textual instructions" that can confuse voters, but the Advantage's user-interface design allows voters to cast their vote without a reminder to vote for all offices and questions listed.

In the New Jersey context, Kimball and Kropf find that in the 2006 general election, for example, three public questions each had residual vote rates between 28 and 30 percent, meaning that nearly one-third of voters failed to correctly cast a vote on each of the public questions. Certainly, this residual vote rate is potentially attributable a range of factors: the active choice not to cast a vote on a specific public question to the bewildering design of the full-face DRE ballot, and, perhaps, even a machine malfunction. Regardless of the cause, however, the fact that almost one in three New Jersey voters did not record a vote on these public questions should be cause for concern, especially if the high residual vote rate is related in some way to the technology being used. Only 24.4 percent of the voting age population turned out to vote in NJ in 2006-and only about a third of that percentage recorded a vote on the public questions. Therefore, only about 8 percent of those eligible to vote actually had a say in the success or failure of the public questions. Does this satisfy our definition of democracy?

Furthermore, a vote cast on a DRE has the potential to be easily altered without the voter's knowledge. In a March 2007 paper (PDF), Andrew Appel, professor of computer science at Princeton University, argues that "it is not difficult to replace the vote-counting software in these machines with fraudulent software that silently shifts votes from one candidate to another without any outward sign to the voter that it is doing so" and proposes implementing voter-verified paper ballots (VVPB) for electronic votes in New Jersey, bolstered by a mandatory hand recount of these paper ballots. With such a system, a voter would be able to review a paper printout of his or her vote before pushing a button to make that vote official. Yet as, Appel notes, in order for paper trails to actually reduce voter error, voters must actually check their vote on paper before officially casting their ballots; obviously, there can be no guarantee that voters will do this.

Most states already have the ability to produce paper records of their votes, but six states will use completely paperless machines for the 2008 elections: Delaware, Georgia, Maryland, Louisiana, South Carolina-and New Jersey. There have certainly been efforts, both on the federal and state levels, to require paper trails. In February 2007, for example, Congressman Rush Holt, who represents New Jersey's 12th district, introduced the Voter Confidence and Increased Accessibility Act (H.R. 811) (PDF), which seeks to mandate the use of a voter-verified paper and to require audits in up to 10 percent of precincts. At present, 216 congressmen have signed on to cosponsor the bill, but no action has been taken on the bill since May of 2007. At the state level, in 2005, the New Jersey state legislature passed a law requiring a voter verified paper audit trail (VVPAT) to backup its electronic voting machines by this November. However, federal inspectors have not yet given the approval necessary for Sequoia to implement paper trails at the manufacturing stage. By the end of last year, a state Senate committee had voted to push back the deadline for meeting the paper trail requirement by six months, in keeping with the recommendation of NJ Attorney General Anne Milgram. The 2008 election, therefore, took place in the state of New Jersey with no paper record of how its citizens voted.

What, then, is the answer to avoiding a potential failure of New Jersey's voting systems? The state has several options that would certainly represent an improvement over the status quo. The state's current machines, however are quite new; fifteen of the state's 21 counties acquired their current DREs after the 2000 election. Abandoning these machines so soon after their purchase would be understandably tough to swallow and thus represents a major obstacle to sweeping change.

State and federal efforts to employ a voter-verified paper trail for electronic voting represent one possible reform. Previously documented issues with paper trails themselves, however, suggest that these efforts may not be enough to produce fair and efficient elections. Printer failure is a possibility-for instance, a paper jam can preclude a voter from seeing a hard copy of his or her vote before (s)he officially decides to cast it. And as already noted, there is no way to ensure that the voter fully inspects the printed paper copy of his ballot. While VVPAT may offer the opportunity for voters to check that their vote has been correctly cast, it is just that-an opportunity.

Another potential solution rests with optical scan technology. According to an October 2008 Election Data Services report (PDF), 1,836 counties (or 56.2 percent of counties) will be using optically scanned paper ballots this year; just eight years ago, only 1,279 counties (29.5 percent) used these ballots. Optical scan ballots operate much like standardized tests-the voter fills in a bubble next to his candidate of choice on his ballot, the ballot is fed through a scanner, and the vote is recorded. This technology could be provided for every polling place in New Jersey at an estimated cost of about $34.5 million, while New Jersey's current DRE system-just the machines themselves, not including storage and transportation costs-has cost around $70 million. (The attachment of printers to these DREs would tack on another $20 million to this cost.) This cost discrepancy is easy to understand, since not only are optical scanners smaller (and therefore easier to store and transport), but only one scanner would be needed per polling place to handle all of the ballots cast there, as opposed to multiple DREs. Sequoia alone has about 10,000 voting machines in New Jersey, and far fewer optical scanners would be needed to ensure efficient elections in the state. A switch from DREs to optical scan machines would certainly not be unprecedented; in February 2007, Florida Governor Charlie Crist announced that DREs would be replaced by optical scan machines in all of the state's precincts. An optical scan voting system would not even be completely new to the state of New Jersey, as optical scan machines are used in 17 New Jersey counties to tally absentee ballots (the four remaining counties instead rely on manual hand counts of these ballots).

Restoring voter confidence in the elections process--and specifically assuring voters that their votes will count as intended--was among the key purposes of HAVA. Yet compliance with HAVA does not necessarily resolve all the issues at hand. Though DREs do allow states to satisfy HAVA's accessibility requirements, their performance can leave a lot to be desired. New Jersey's experience with DREs is indicative of various problems with the machines, including their high costs and high residual rates.

What is to be done, then? In the short term, New Jersey would be well-served to implement a reliable paper backup for its electronic votes before the next statewide election. But simply having a paper printout is no guarantee that voters will actually check to see if their vote has been recorded correctly; a 2006 Cuyahoga County, Ohio report (PDF) found that many voters are not even aware that there is a printout to check. This issue of voters not checking their ballots might be ameliorated with an easier-to-read ballot that makes it clear to voters that there is a printout to be examined before their vote becomes official. Additionally, there could be statewide mailings informing voters of the new printer systems so the idea of a printout is not novel to voters when they enter the voting booth. In the long term, it may be advisable for New Jersey to make the switch from DREs to optical scan. However, the ultimate decision to make this switch would depend upon the performance of the DREs with printers. If DREs with printers result in fewer residual votes, then a change to optical scan may be deemed unnecessary. Instead, more incremental change to the DREs may be suitable.

Jennifer Marsico can reached at jennifer.marsico@aei.org.

Viewpoint is an occasional feature analyzing various election reform issues.